A thousand-name text. Eleven of the names are missing, because a scan broke across two pages.
Nobody would ever have checked. It is name 707 through 717 of a thousand — a stretch of epithets in a devotional list, in a genre where epithets are largely interchangeable and where inventing eleven plausible ones would take an afternoon and be undetectable forever.
The rendering printed this instead, in the name-slot itself:
GAP IN SOURCE: the scan breaks mid-sentence at the end of p.606 ("amoghaḥ…") and resumes mid-gloss on p.607. Names 707–717 — eleven names and the Sanskrit verse containing them — are not recoverable from this source; NOT FABRICATED HERE. Flagged for anyone with access to a second print copy.
It locates the break to the page. It quotes the broken text on both sides. It says what would close it. And it states, unprompted, that nothing was invented.
I have spent ten phases cataloguing the opposite move. A corpus that manufactures chains of custody it does not have. Quotations attributed to a volume that does not contain them. Five obstacle classifications credited to a source that never lists them. Provenance frames applied to the author's own coinages at a rate of about thirty-five to one in a single document.
Against all of that: a source that declined to manufacture content it could trivially have invented, and said so where the content would have gone.
And here is the thing I did not expect. The hole raised my confidence in the other 989 names more than a complete list would have. A seamless text of unknown provenance tells you nothing. A text that shows you exactly where it failed has told you the shape of its own reliability — and has implicitly certified everything it did print.
That is a real epistemic asset and it costs nothing but nerve.
As a house rule for the book: every gap gets a slot. Not a footnote, not an acknowledgements-page caveat about the limits of the research — a marker in the position where the missing thing belongs, saying what is absent, how far the absence extends, why it is absent, and what would repair it.
The temptation runs the other way, and it is strong precisely where the material is weakest. A chapter with three unsourced constructs reads better if the three are asserted smoothly. It is more trustworthy if each carries a slot saying: this one is mine, I could not find it in the tradition, here is where I looked.
The corpus's own best formal device already does half of this — the two-line conversion table, Traditional: above Daily Life Equivalent:, which makes a swap visible instead of silent. The gap-slot is the same discipline applied to absence rather than to substitution.
⚑ Both are the same underlying move: show the seam. A reader who can see where you joined things can trust the joins. A reader who cannot has to trust you, and has no reason to.